Accurate diagnosis of thyroid cancer using a combination of surface-enhanced Raman spectroscopy of exosome on MXene-coated gold@silver core@shell nanoparticle substrate and deep learning

Exosomes (EVs), serving as one of the optimal subjects for liquid biopsies, have seen broad applications in diagnosing various diseases, including cancers. In this paper, an efficient method has been proposed for label-free profiling of exosomes in biological samples (e.g., plasma) by a combination of surface-enhanced Raman spectroscopy (SERS) on MXene-coated gold@silver core@shell nanoparticle (Au@Ag NP) functionalized substrate and deep learning. Due to the contributions of electromagnetic enhancement (EM) and chemical enhancement (CM) of MXene-coated Au@Ag NP substrate, the as-proposed SERS sensing platform exhibits a dynamic range of 0.5 × 1010 to 2.0 × 1011 EVs mL−1 with a limit of detection (LOD) as low as 1.7 × 109 EVs mL−1 (three times standard deviation (3σ) of blank sample). Subsequently, a deep-learning classification algorithm has been developed for extracting the features of EVs from complex Raman spectra by residual neural networks. As a proof of principle, the preliminary validation of our approach is demonstrated by discrimination of thyroid cancer patients from healthy controls with diagnostic accuracy of 96.0 %, and staging of the cancer patients with accuracy of 86.6 %, respectively.